Vehicle Recognition via Parallax Strip Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle recognition systems struggle to individually recognize vehicles that overlap with each other in photographed images, often incorrectly detecting white lines or other road features as vehicles.
Innovation Solution
A vehicle recognition apparatus using a stereo camera to generate a parallax image, divide it into strip regions, calculate representative parallaxes, extract efficient strip regions, couple adjacent regions based on a distance distribution model, and recognize vehicles by grouping coupled strip regions, while distinguishing between vehicles and non-vehicle regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a stereo camera is used to photograph vehicles on a road, then the nearest vehicle can be easily recognized, but it becomes difficult to individually recognize vehicles that overlap with each other in the photographed image
Solution Approach 1:
The photographed image is divided into multiple strip regions along the vehicle width direction. Each strip region is processed independently to extract depth information and vehicle features, allowing overlapping vehicles to be separated and recognized individually rather than as a single merged object
Solution Approach 2:
The patent utilizes the depth dimension by generating a depth map from stereo camera images. By adding this third dimension (depth) to the two-dimensional image data, vehicles at different distances can be distinguished even when they overlap in the image plane, enabling individual recognition of overlapping vehicles
2Device complexity
If conventional image processing is used to detect vehicles, then processing is simple, but road features such as white lines are erroneously detected as vehicles
Solution Approach 1:
The image is segmented into strip regions, and each region is analyzed for depth consistency and vehicle-specific features. This segmentation allows the system to distinguish between road features (which typically show consistent depth across strips) and actual vehicles (which show depth variations characteristic of three-dimensional objects), reducing false detections while maintaining processing efficiency
Solution Approach 2:
A depth map is introduced as an intermediary data structure between the raw image and vehicle detection. This depth information acts as a mediator that provides additional evidence to distinguish vehicles from road features, improving detection reliability without significantly increasing processing complexity
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate individual recognition of overlapping vehicles, reducing erroneous detections and improving the counting of vehicles in a column, by effectively differentiating between vehicles and road features.
Implementation Method 1
generate a parallax image based on the photographed image
Data Source
AI summary
A vehicle recognition apparatus according to an embodiment includes one or more hardware processors to: acquire a photographed image obtained by photographing vehicles forming a column of vehicles from an obliquely rearward direction with respect to the column of vehicles; generate a parallax image based on the photographed image; estimate a vanishing point in the parallax image; generate strip regions by dividing a region below the vanishing point in the parallax image in a horizontal direction; calculate a representative parallax for each strip region; extract strip regions efficient in recognizing a vehicle from among the strip regions based on the representative parallax; couple strip regions adjacent to each other among the extract strip regions to generate a coupled strip region based on a distance distribution model of a back surface and a side surface of a vehicle; and recognize a vehicle based on the coupled strip region.